Onion

onion (ONe Instance ONly) is a tool for removing duplicate parts from large collections of texts.

Paper | Cite | Licence

Installation

Prerequisites

• 64-bit CPU architecture
• Google sparse hashset (​https://github.com/sparsehash/sparsehash)

Configuration and installation

1. Download the sources:

   wget -O onion-1.4.tar.gz 'http://corpus.tools/raw-attachment/wiki/Downloads/onion-1.4.tar.gz'
   

2. Extract the downloaded file:

   tar xzvf onion-1.4.tar.gz
   

3. Configure the package by editing onion-1.4/Makefile.config:
   • set PREFIX (or INSTALL_BIN and INSTALL_DATA) according to where you want the executables and data (docs) installed
4. Install the package (you may need sudo or a root shell for the last command):

   cd onion-1.4/
   make
   make install
   

Quick start

onion -s <documents.vert >deduplicated_documents.vert

There's also an usage example on a sample input.

For usage information see:

onion -h
man onion

Usage

onion [OPTIONS] [FILE]

Mark duplicate text parts in the input vertical file.

 -f FILE   hashes of duplicate n-grams
 -n NUM    n-gram length (default: 5)
 -t NUM    duplicate content threshold (default: 0.5)
 -d STR    document tag (default: doc)
 -p STR    paragraph tag (default: p)
 -s        strip duplicate parts (rather than mark)
 -m        no smoothing
 -T NUM    trim n-gram hashes to NUM bits (default: 64)
 -l NUM    max stub length (default: 20)
 -b NUM    buffer size, in bytes (default: 16777216)
 -q        quiet; suppress all output except for errors
 -V        print version information and exit
 -h        display this help and exit

With no FILE, or when FILE is -, read standard input. Output is written to standard output.

Acknowledgements

This software has been developed at the ​Natural Language Processing Centre of ​Masaryk University in Brno with financial support from ​PRESEMT and ​Lexical Computing Ltd. It also relates to Jan Pomikálek's ​PhD research.

Licence

Onion is licensed under the ​BSD 3-Clause License